which analysis to use?

lucyo

New Member
#1
which analysis to use? MANOVA?

Hi guys,

I'm really struggling with choosing the right statistical analysis to interpret my data for my MSc, any help would be really appreciated!

I have a repeated measures design looking into the effects of recognising visually degraded words that are either food related or non-food related items when participants are either hungry or satiated. My hypothesis is that participants will identify the food related items faster than the non-food related items when they are in the hungry condition than when they are in the satiated condition. I want to also see if there is an interaction with high calorie content foods being identified faster than low calorie content foods, and if there is a difference for males and females. The dependent variables are the amount of time taken to identify the word, and the percentage of word that needs to be revealed to identify it.

This has confused me for a couple of reasons... The first is that two of my IVs overlap, i.e. food items vs non-food items and then high calorie food items vs low calorie food items. Should I combine these into one IV: high calorie vs low calorie vs non food items?

The second is that my DVs of time taken to identify word and percentage revealed to identify it will refer to all of the words, when I want to see if there is a difference between these values for food items and non food items separately. Does this mean I have 4 DVs? i.e. time taken to identify food items, percentage revealed to identify food items, time taken to identify non food items, and percentage revealed to identify non food items.

I've really confused myself here and am thinking I might need a MANOVA? If any one can help me see this a bit clearer I would really appreciate it!!

Please reply, even if you don't think it will help that much, any help is welcomed!!

Lucy
 
Last edited:

terzi

TS Contributor
#2
Hi Lucy,

I hope to be helpful, but you'll let me know :D. Since you have only two dependent variables, I would use two ANOVA models, one for each DV, in order to avoid using MANOVA, which requires a deeper understanding on multivariate statistics. First I would search for a difference between food related items and non-food items. If the difference is significant, you could merge the IV to create one with three categories and look for differences among those three treatments (Maybe non-food items and low calorie items have same response times, only high calorie ones are different)
 

lucyo

New Member
#3
Thank you for your response! you've help me see that maybe it would be too tricky to try to analyse it all in one go. I'm still a little confused though, when you say one ANOVA for each DV do you mean the DV of percentage of food related word revealed then another ANOVA for percentage of non-food related items revealed? So one ANOVA of 2 (state; hungry vs satiated) X 2 (gender; male vs female) with a DV of percentage needed to be revealed to identify food related items, hoping to find a main effect of state, and perhaps an interaction with gender? (research indicated it might affect females more than males). Then another ANOVA doing the same but with a DV of percentage needed to be revealed to identify non-food related items? Hoping to find no significant effects? Then to repeat the first ANOVA but with a different layout, like a 3 (high calorie vs low calorie vs non food items) X 2 (male vs female) ANOVA with a DV of percentage revealed? But then then DV of percentage revealed will mean for all words... I think in my DV I need to specify whether it's food or non food word percentage revealed don't I? So I could do a 2 (high vs low calorie) X 2 (males vs female) with a DV of percentage of food words needed to be revealed?

Does that make sense? The more I think about it, the more confused I get! Am I now trying to use ANOVA when it's not appropriate?

Thanks again for your reply, I really appreciate it!
 

terzi

TS Contributor
#4
Hi Lucy,

I understood that you have two DV:

1.time taken to identify word
2.percentage revealed to identify it

My suggestion is to make two ANOVAS first, using each one of your DV, comparing food related items and non-food items. You can also include gender in those models, as a second independent variable.

Assuming you get a significant result, that would mean that there is a difference between food related items and non-food related. You then use another two ANOVAS, one for each DV, and consider only the difference between low calorie and high calorie items.

Actually you could do the analysis using only two ANOVAS, one for each DV and using your combined IV (i.e. high calorie vs low calorie vs non food items) and gender in the same model. But if you don't know much about statistics that would be more complicated since it would require post-hoc tests, and interpreting interactions can be more complicated.